Over the past week, a large part of our focus shifted toward one of the more practical and high-impact use cases we’re currently working on: real estate.

A key step there has been expanding how Nexus handles unstructured data. We can now fully query PDF-based datasets containing complex, unstructured information directly inside Nexus, turning it into structured and actionable data. That alone already changes how teams can work with fragmented information, but this week we pushed it further.

For the first time, we successfully ingested hundreds of lengthy phone conversations (audio), transcribed them into structured and actionable data, and made both the summaries and extracted actions queryable through Nexus itself. That means conversations are no longer isolated pieces of information sitting in recordings or transcripts, but become part of a broader operational knowledge layer that teams can actually interact with.

What makes this particularly important is the performance and quality level we’re seeing. The results are reaching a point where real-time processing becomes a realistic next step. That opens the door toward much broader applications around call center intelligence, AI agents, and systems capable of suggesting “next best actions” during live interactions. It’s an area receiving a lot of attention across the market, and seeing these capabilities already working within Nexus is a significant milestone.

On the business side, we had an important follow-up call with the PoC customer in Qatar. The focus is now shifting toward final delivery and evaluating the potential next phase of the process. Much of the work happening behind the scenes right now is about making sure the product holds up under real operational conditions.

Prioritization

Internally, one of the strongest themes this week has been prioritization.

There is a lot happening simultaneously across Nexus, client projects, integrations, and new opportunities. The risk in that situation is ending up working on everything at once while not moving anything forward meaningfully. Because of that, we’ve become increasingly focused on a simple question: what delivers direct value right now, either for Nexus itself or for the clients actively using it?

That approach means some features are being developed that weren't formally listed on the roadmap. Instead, they emerged directly from active PoCs and customer workflows where solving a concrete problem mattered more than following a predefined feature sequence. In practice, this has proven to be a much stronger driver for product maturity.

That prioritization is also directly influencing how we sequence development. As a result, capabilities such as speech-to-text processing, expanded RAG functionality, and several new and improved connectors have been prioritized in the planning, while other areas, such as the Google Marketplace rollout, have temporarily slowed down. Another example can be found in the group chat agents. The existing setup was not yet performing at the level we wanted, especially when it came to instruction clarity and coordination between agents. Because of that, we prioritized features that allow much more specific upfront instructions for individual agents, while temporarily pausing parts of the public workspaces feature.

Solude

Alongside the broader Nexus developments, we also had an alignment session with Solude. Over the coming weeks, the focus will shift toward further testing of the current integration and identifying where Nexus-driven processes can create operational advantages within the Solude platform itself.

The ambition there goes beyond simply controlling a platform through a chat interface. The real goal is identifying friction points, operational bottlenecks, and repetitive workflows where AI-native orchestration can meaningfully improve how work gets done.

Overall, the current phase continues to revolve around turning real-world usage into product direction. Whether it’s dashboards, document exports, multi-agent orchestration, or now conversational intelligence, the common thread remains the same: building systems that are shaped by actual operational needs.